srclight

Deep code indexing MCP server for AI agents. 25 tools: hybrid FTS5 + embedding search, call graphs, git blame/hotspots, build system analysis. Multi-repo worksp

What is srclight?

Deep code indexing MCP server for AI agents. 25 tools: hybrid FTS5 + embedding search, call graphs, git blame/hotspots, build system analysis. Multi-repo workspaces, GPU-accelerated semantic search, 10 languages via tree-sitter. Fully local, zero cloud dependencies.

Overview of srclight

What it is

Srclight is a deep code indexing MCP server for AI agents. It builds a searchable index of a codebase using SQLite FTS5, tree-sitter, and embeddings. It provides 42 tools covering symbol search, relationship graphs, community detection, impact analysis, git change intelligence, semantic search, build system awareness, and document extraction. It is fully local and private, with no cloud dependencies. It supports 19 languages via tree-sitter and 10 document formats. It works with Claude Code, Cursor, and any MCP client. It also has a CLI for indexing, searching, and inspecting from the terminal. It supports multi-repo workspaces via SQLite ATTACH+UNION. It offers optional OCR via PaddleOCR and pytesseract. It provides 4 search modes: symbol names, source code (trigram), documentation (stemmed), and semantic (embeddings). It uses hybrid search with RRF fusion of keyword and semantic results. It supports incremental indexing with content hash detection. It has auto-reindex via git post-commit/post-checkout hooks. It provides index freshness verification and index coverage reporting. It uses MCP argument validation that refuses unknown tool arguments. It provides graph resolution labels for caller/callee edges. It requires Python 3.11+, Git, and optionally Ollama for semantic search, NVIDIA GPU + cupy for GPU-accelerated vector search, and Poppler for PaddleOCR.

Key features

  • 42 tools covering symbol search, relationship graphs, community detection, impact analysis, git change intelligence, semantic search, build system awareness, and document extraction.
  • Hybrid search with RRF fusion of keyword + semantic results.
  • Multi-repo workspaces via SQLite ATTACH+UNION.
  • Incremental indexing with content hash detection.
  • Auto-reindex via git post-commit/post-checkout hooks.
  • Index freshness verification and index coverage reporting.
  • MCP argument validation that refuses unknown tool arguments.

Use cases

Srclight is used by AI coding agents (such as Claude Code and Cursor) to reduce token waste on code orientation. It replaces multiple grep/glob calls with precise, structured lookups. It can be used for finding callers, understanding module structure, performing semantic search, detecting change impact, and reducing tool calls per bug fix.

Getting started

See the repository README for installation and quick-start commands.

License and pricing

The license is MIT. The pricing type is open source.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/srclightsrclight.svg)](https://yunodb.com/tools/srclightsrclight?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/srclightsrclight?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/srclightsrclight.svg" alt="Featured on YunoDB" height="20"></a>